def test_AIC_score(self):
y_true = [1, 2, 3, 4, 5]
y_pred = [1, 3, 2, 5, 4]
self.assertAlmostEqual(
AIC_score(y_true, y_pred, MockModel(coef_=[1] * 2)),
21.0736, places=3)
self.assertAlmostEqual(
AIC_score(y_true, y_pred, df=3),
AIC_score(y_true, y_pred, MockModel(coef_=[1] * 2)), places=3)
self.assertGreater(
AIC_score(y_true, y_pred, df=3),
AIC_score(y_true, y_pred, df=2))
def display_logistic_regression(df, train_x, valid_x, train_y, valid_y):
print('(8) Display logistic regression\n')
# fit a logistic regression (set penalty=l2 and C=1e42 to avoid regularization)
logit_reg = LogisticRegression(penalty='l2', C=1e42, solver='liblinear')
logit_reg.fit(train_x, train_y)
print('intercept ', logit_reg.intercept_[0])
print(
pd.DataFrame({'coeff': sorted(abs(logit_reg.coef_[0]), reverse=True)},
index=PREDICTORS), '\n')
print(
'AIC',
AIC_score(valid_y,
logit_reg.predict(valid_x),
df=len(train_x.columns) + 1))
classificationSummary(train_y, logit_reg.predict(train_x))
classificationSummary(valid_y, logit_reg.predict(valid_x))
prediction_valid = logit_reg.predict(valid_x)
prediction_train = logit_reg.predict(train_x)
print('precision on test is:', precision_score(valid_y, prediction_valid))
print('recall on test is:', recall_score(valid_y, prediction_valid))
print('f1 on test is:', f1_score(valid_y, prediction_valid))
print('Logistic Regression:Accuracy on train is:',
accuracy_score(train_y, prediction_train))
print('Logistic Regression:Accuracy on test is:',
accuracy_score(valid_y, prediction_valid), '\n')
def score_model(model, variables):
if len(variables) == 0:
return AIC_score(y, [y.mean()] * len(y), model, df=1)
return AIC_score(y, model.predict(X[variables]), model)
def score_model(model: LinearRegression, variables: list):
if len(variables) == 0:
return AIC_score(y, [y.mean()] * len(y), model, df=1)
return AIC_score(y, model.predict(x[variables]), model)
def score_model(model, variables):
if len(variables) == 0:
return AIC_score(train_y, [train_y.mean()] * len(train_y), model, df=1)
return AIC_score(train_y, model.predict(train_X[variables]), model)
def score_model(model, variables):
return AIC_score(train_Y, model.predict(train_X[variables]), model)
axes[1].set_title('validation')
axes[2].set_title(' ')
axes[2].set_ylim(-30000, 30000)
plt.suptitle('Prediction errors')
plt.subplots_adjust(bottom=0.15, top=0.85, wspace=0.35)
plt.show()
# In[27]:
#Adjusted R2 BIC and AIC
pred_Y = reg.predict(train_X)
print('adjusted r2 : ', adjusted_r2_score(train_Y, pred_Y, reg))
print('AIC : ', AIC_score(train_Y, pred_Y, reg))
print('BIC : ', BIC_score(train_Y, pred_Y, reg))
# Use predict() to make predictions on a new set
reg_lm_pred = reg.predict(valid_X)
result = pd.DataFrame({
'Predicted': reg_lm_pred,
'Actual': valid_Y,
'Residual': valid_Y - reg_lm_pred
})
print(result.head(20))
# Compute common accuracy measures
regressionSummary(valid_Y, reg_lm_pred)